Follow-the-Perturbed-Leader for Adversarial Markov Decision Processes with Bandit Feedback.
Yan DaiHaipeng LuoLiyu ChenPublished in: NeurIPS (2022)
Keyphrases
- markov decision processes
- optimal policy
- state space
- reinforcement learning
- finite state
- planning under uncertainty
- policy iteration
- transition matrices
- reinforcement learning algorithms
- partially observable
- reachability analysis
- dynamic programming
- average reward
- decision processes
- infinite horizon
- average cost
- risk sensitive
- factored mdps
- finite horizon
- markov decision process
- decision theoretic planning
- model based reinforcement learning
- stochastic shortest path
- reward function
- partially observable markov decision processes
- sufficient conditions
- state and action spaces
- markov chain
- linear programming
- multi agent
- real time dynamic programming
- decision making